First of all, a big thank you to everyone who completed the survey on the state and impact of Fortran. Thank you also, again, to the those of you who piloted an earlier version of the survey. That piloting helped significantly with the design, and therefore the effectiveness, of the survey.
Second, I appreciate that people will also want to see the results. My colleagues and I have started to analyse the data. As a token of our thanks I’ll share a summary below of some emerging findings. As we progress with the analysis, and prepare publications, I will share further information. We are committed to open sourcing the dataset, though this may take some time, e.g., we want/need to publish some of this first.
Third, and related to publications, I know there is interest amongst those in the Fortran community to publish a response to the LANL Report on Fortran, with at least one survey respondent expressing interest in co-authorship (I will contact them). If others are interested in being involved, in some way, in preparing a response, please let me know (message me on discourse, or email me at a.rainer@qub.ac.uk). To reassure you, I’ll treat an approach from you as an expression of interest at this stage, with no expectation of commitment.
Fourth, to try to ensure I get this thank you message out to as many people as possible, I’ll be sending an email, sometime in the next few days, so you may get this message again, via another channel.
Finally, I provide a ‘peek’ at the responses below. I enumerate them for ease of cross-reference, rather than for priority. Much of the following is descriptive, but that’s still valuable because I am unaware of any prior attempt to characterise the global Fortran community and, as we found out in the UK, the community is mostly invisible, even to itself. Thus, the following descriptive information helps the community to better understand and appreciate itself.
Here goes:
- We had 150 responses. All responses were high quality, an indication that everyone who responded did so conscientiously. That really helps with the rigour of the results.
- 27 countries are represented in the responses (26 if one excludes the respondent that identified their codebase as located on “The Moon”). The majority of responses came from the UK (33%), mainland Europe (36%, of which Germany was 13% and France was 8%) and the USA (20%). The high response rate for the UK may be because, in recent years, we have worked to bring together the UK Fortran community.
- Most respondents are from universities or similar (37%), the open source community (23%) or government (16%) though 23 respondents (15%) were from industry. This gives an indication of where Fortran continues to have an impact. More on this below.
- There was a real mixture of roles, which we have (for now) aggregated into: academic (33%), Research Software Engineer or similar (25%), computational scientist (23%) and volunteer (9%). Some respondents did question the difference between academic and computational scientist. That may be a weakness in the design of the study.
- There was a fairly uniform spread of respondents’ age: 25yrs - 34yrs (20%), 35-44 (27%), 45-54 (20%), and then a tail off. This spread of ages is encouraging for the future of the Fortran community.
- In terms of gender: male (87%), female (8%), non-binary (1%). This distribution is unfortunate but also not unexpected.
- 80% of respondents had doctorates, which is at least partly explained by the fact that Fortran is mostly used in research-related contexts. The majority (82%) of respondents, irrespective of specific qualification, said that their qualification was essential or directly relevant to the work they did.
- Most respondents (79%) work on more than one Fortran codebase. One implication is that, in my view, we made the correct survey-design choice, i.e., to ask respondents to focus on a single codebase, and thus reduce the “it depends” responses. That said, that most respondents work on more than one codebase is an indication that there may be many codebase we have “missed” in the survey. Of course, it’s hard to know what effect that survey-design choice has had - something for us to explore in due course.
- Respondents chose the single codebase for many reasons, the most frequent being: the codebase I most frequently work on; the most important codebase for me; the most significant codebase to science and engineering. On this final point, see item 15 below.
- The majority of codebases chosen were applications (75%), with a sixth (16%) being libraries, and then tools and other types of codebase.
- 58 respondents (39%) provided URLs for git-based repositories, the majority being github. Some respondents provided URLs to their own domain, which may be git-based (it’s not obvious to tell from the URL). Others provided the name of the codebase. Others explained the codebase is private. URLs are of course helpful for us, e.g., to undertake further analyses of the codebases to better characterise codebases.
- A 1/4 of codebases are developed and maintained by one person, with 2/5ths (40%) developed and maintained by 2 - 5 people. In other words, almost 2/3rds of codebases are maintained by small teams. This of course raises risks about the future of these codebases.
- Codebases varied in size from ~1000 LOC to 5MLOC to some indication of more than that. 50% of codebases are considered to be large or very large.
- 76% of codebases are not considered to be legacy codebases by the respondents. A very similar proportion (~70%) occurs in relation to others, i.e., 70% of respondents thought that others do not consider the codebases legacy either. Recognising that there may be an dependence between these %s, the %s are interesting because of the perception that people outside of the Fortran community have about Fortran, i.e., that Fortran and Fortran codebases are legacy codebases.
- In terms of the scientific and engineering disciplines ‘dependent’ on the respective codebases, there’s more analysis we need to do but the following extracted keywords give an indication of the disciplines - the square brackets provide context: astronomy, fluid [dynamics’], CFD, flow, climate, ecosystem, weather, environment, earth, water, magnetics, chemical, chemistry, physic [stemmed to cover physical and physics], particle, quantum, X-ray, plasma, matter, material, mechanical, fusion, nuclear, aero [aerospace], automotive, engineering, genetic, electronic, hydraulics, crystallography, economic, health, data, PDE, signal [processing].
- Finally, the survey asked respondents how they’d like to be acknowledged. Where real names were given for acknowledgement, with no alternatives provided, we have retained those names (though removed email domain names). On a few occasions, we have ‘played safe’ at the moment and not named individuals. We acknowledge: an RSE from a Helmholtz research institute in Germany, “Keeping people and society safe”, AdAstraPerAspera, Amir Shahmoradi, Amy Roberts, Anon77, ariiJudeIS, BiathlonFan, Boteh, CabraLattice, CampanologicalToxophilite, CampFire, CJM, CLaSSGroupLead, cmaapic , CoolFortran, Dalekopera, dmio7839r, DogCatMouse, DrFortranFan, dwa, ForProfit, FortranIsStillMyFavorite, FortranKPDH, Fotranista123BEva, FragmentedEchidna, Gaits, Gideon, GOTO_EndOfSurvey, GoTo_remover, HerbertPeck, hjf, ibperth, IndAtHomeFort18, InTartifletteWeTrust, jasNW, lsm, Machalot, Martin Diehl, mathemagician, Milan Skocic, Modeller2026, mwatkins, NumberCrunching4QD, OldCrustyAstroGuy, otisthecat, paleoprogrammer, Partecipant 21, ParticipantJad, PDS, PH, Phoenix, platipodium, Polis, PRIMA, RSEwithAGPL, SelaBuri, Simulacrum, T3, TNT, upg88743, yarchik.
There is of course more - much more - analyses to follow, both in terms of questions that we’ve not looked at yet (e.g., licenses, Fortran revisions, challenges), but also in terms of sub-samples (e.g., how the results for applications compare with those for libraries, how results for Research Software Engineers compare with those of academics/scientists).
If you’ve any questions, please either ask them here on the discourse space, message me direct on discourse, or email me.
Thank you again to everyone who contributed to the survey.
Austen
). I am hoping that the following link will work:
Thank you.
. The survey closes on 12 Feb. If you’d like to complete the survey and have not done so yet, there’s still a couple of days remaining. We have 123 responses so far, which I am personally very pleased about.
